search

nyu-mll / jiant-v1-legacy

The jiant toolkit for general-purpose text understanding models
MIT License
21 stars 9 forks source link

[CLOSED] WiC and winogender-diagnostic - TypeError: apply_standard_boundary_tokens() got an unexpected keyword argument 'get_offset' #1104

Closed jeswan closed 4 years ago

jeswan commented 4 years ago

Issue by hsajjad Monday Jul 20, 2020 at 14:30 GMT Originally opened as https://github.com/nyu-mll/jiant/issues/1104

Hi, I am getting the following error when training/finetuning on wic and winogender-diagnostic. The rest of the tasks are processed without a problem. Is there something that I am missing?

07/20 04:35:33 PM: Task wic (train): Indexing from scratch. 07/20 04:35:33 PM: Fatal error in main(): ... TypeError: apply_standard_boundary_tokens() got an unexpected keyword argument 'get_offset'

I am using the jiant/config/tutorial.conf

Thanks, Hassan

jeswan commented 4 years ago

Comment by sleepinyourhat Monday Jul 27, 2020 at 16:17 GMT

Hrm. @W4ngatang, @HaokunLiu, @zphang: Do you remember seeing this in any of our SuperGLUE runs?

jeswan commented 4 years ago

Comment by HaokunLiu Monday Jul 27, 2020 at 16:23 GMT

I added the argument for transformer based input models, but I forgot to address the older models, these tasks probably only work for transformers now.

If using an older model is necessary, adding a dummy argument to the function mentioned the error message should fix the problem.

On Mon, Jul 27, 2020 at 12:18 PM Sam Bowman notifications@github.com wrote:

Hrm. @W4ngatang https://github.com/W4ngatang, @HaokunLiu https://github.com/HaokunLiu, @zphang https://github.com/zphang: Do you remember seeing this in any of our SuperGLUE runs?

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub https://github.com/nyu-mll/jiant/issues/1104#issuecomment-664495286, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIPK5GSIQY4UDCRCSGMCRCLR5WSEBANCNFSM4PCMOKIQ .

jeswan commented 4 years ago

Comment by hsajjad Monday Jul 27, 2020 at 16:38 GMT

I am trying to reproduce the glove numbers reported in the paper so you are right that I am not using transformer based input models. I can try adding get_offset to the function apply_standard... if that's the right way to fix it